Judgment Playbook

Two-way / one-way door

Match decision speed and evidence depth to how costly the choice is to reverse.

5 minute readtaught in course chapter 8
UncertaintyCommitment

What it is

The door test sorts a decision by how hard it is to undo. A two-way door can be walked back once you have learned something: you reopen it and return to where you were. A one-way door closes behind you, because it moves capital, signs contracts, spends trust, or changes what other people do next. Reversibility, not size, sets how much evidence and deliberation the choice deserves.

Where it comes from

The door metaphor is Jeff Bezos's, from Amazon's letter to shareholders for 2015, published in April 2016. He wrote that "some decisions are consequential and irreversible or nearly irreversible – one-way doors – and these decisions must be made methodically, carefully, slowly, with great deliberation and consultation." Most decisions, he continued, "are changeable, reversible – they're two-way doors" and "can and should be made quickly by high judgment individuals or small groups." He called the first kind Type 1 and the second Type 2.

The economics behind it is older and has no single author. Kenneth Arrow and Anthony Fisher argued in 1974 that when a harm cannot be undone, uncertainty on its own is a reason to wait for information; they named the value of keeping the choice open quasi-option value. Avinash Dixit and Robert Pindyck built that into the standard account of irreversible investment in 1994. The economists supplied the reason, Bezos the name a team can use in a meeting on Monday.

What it corrects

Large organisations converge on a single decision process, and the one that survives is the one that protects people from blame: slow, consultative, heavy on evidence. It then gets applied to choices that could be tested in a fortnight and undone in a day. The cost never appears in a report, because it consists of experiments not run. Ordinary care cannot fix this, since care is the thing being over-supplied. Bezos put the asymmetry in a footnote: firms that habitually use the light process on Type 1 decisions "go extinct before they get large," so the survivors are the over-careful ones. Fast companies flatten the same way in the other direction, giving a door that locks behind them the speed the team is proud of.

How it works

  1. Write the decision as the specific first move, not as the strategy behind it.
  2. Split that move into the part that can be undone and the part that cannot.
  3. Price the reversal of each part: the weeks, the money, and the people whose agreement you would need to stand where you stand today. A part with no nameable path back counts as one-way.
  4. Shrink the first move until the one-way part is as small as the learning allows.
  5. Match the process to what is left: few people and little delay for the reversible part, deliberation and dissent for the door that stays shut.
  6. State the evidence that would justify walking through the one-way door later, and the date you will look for it.

Worked example

Netflix in 2011 made two decisions that are usually remembered as one. In July it separated DVD-by-mail from streaming into two plans at $7.99 each. Its letter to shareholders of 25 July 2011 gave the effect plainly: for members who wanted both, the change was "as much as a 60% increase for them from $9.99 to $15.98." In September it announced Qwikster, the DVD service rebranded on a separate site with separate billing. On 10 October, about three weeks later, Netflix dropped Qwikster and kept one site.

Splitting the two is the lesson. The rebrand was a two-way door: a name, a website, a queue, all reversible, and duly reversed. The repricing was not. The quarterly letter of 24 October 2011 treated them as different kinds of thing. The branding episode "caused a temporary cancellation surge" whose impact was "relatively minor." The price change had "greatly upset many domestic Netflix members" and had "hurt our hard-earned reputation, and stalled our domestic growth." Domestic subscribers fell from 24.59 million to 23.79 million, the first drop in the two-year run of quarters printed in that letter. Netflix told investors it was "done with pricing changes." The cheap door reopened in three weeks. The expensive one stayed shut.

In a Business Case Weekly case

The Sega case stops in January 2001, with the Dreamcast selling well after a price cut and the PlayStation 2 on shelves. The fork is how long to keep building the console. The door test does not settle it; it splits it. Another quarter of production is spending that can be stopped next quarter. Announcing the end of Sega hardware releases the developers, the retail space and the launch slots a platform is made of, and no later change of mind buys them back. Price the two separately before you argue for either.

In your answer

  • "The reversible part of this decision is …, and we could undo it in … weeks for about …"
  • "The part that closes behind us is …, because it changes … for people outside the company."
  • "So I would move quickly on … and hold … until … is true."
  • "If I am wrong, the path back is …, it needs … to agree, and it costs …"

Common misuse

The phrase becomes a permit. "It's a two-way door" gets said about a move that is reversible on paper and not in the world: the pilot a customer will plan a roadmap around, the price test that teaches a market what your product is worth. Reversible is a measurement, not a mood. The test is one question asked out loud: who has to agree to the reversal, how long does it take, what does it cost? A move whose path back you cannot describe is a one-way door being called a two-way one.

References

Fragility audit

Next method

Find the dependencies that can destroy the plan before trying to predict the exact shock.

Open next tool